Head recognition from depth image

• US 9,081,999 B2
• Filed: 12/28/2010
• Issued: 07/14/2015
• Est. Priority Date: 12/28/2009
• Status: Active Grant

• Current Assignee: SoftKinetic Software
• Original Assignee: SoftKinetic Software
• Inventors: Baele, Xavier, Thollot, Julien, Martinez Gonzalez, Javier, Simons, Kevin
• Primary Examiner(s): DRENNAN, BARRY T
• Application Number:
  US13/519,502
• Publication Number:
  US 20130022262A1
• Time in Patent Office: 1,659 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06V 20/64 Three-dimensional objects
• CPC Class Codes: G06V 40/162 using pixel segmentation or colour matching

Abstract

Described herein is a method for recognizing a human head in a source image. The method comprises detecting a contour of at least part of a human body in the source image, calculating a depth of the human body in the source image. From the source image, a major radius size and a minor radius size of an ellipse corresponding to a human head at the depth is calculated, and, for at least several of a set of pixels of the detected contour, generating in an accumulator array at least one segment of an ellipse centered on the position of the contour pixel and having the major and minor radius sizes. Positions of local intensity maxima in the accumulator array are selected as corresponding to positions of the human head candidates in the source image.

Claims

1. A computer-implemented method for recognising a human head in a source image, the source image having depth values for each pixel thereof, the method comprising the steps of:

a) segmenting at least part of a human body from a background in said source depth image;

b) detecting edges of said segmented part of a human body to determine a contour of said part; and

c) determining the position of a human head within the source image;

characterised in that step c) further comprises the steps of;

—

c1) calculating a depth of said part of a human body in said source image;

c2) calculating a major radius size and a minor radius size of an ellipse corresponding to the size of a human head at said calculated depth in said source image;

c3) generating an accumulator array for at least one segment of an ellipse centred on the positions of a set of contour pixels of said contour, said at least one segment of said ellipse having said major and minor radius sizes; and

c4) selecting a position of a local intensity maximum in said accumulator array that corresponds to the position of a human head in the source image.

2. A computer-implemented method according to claim 1, wherein said depth is an average depth of said part of a human body in said source image.

3. A computer-implemented method according to claim 1, wherein said depth and said major and minor radius sizes are individually calculated for each contour pixel.

4. A computer-implemented method according to claim 1, further comprising the step of carrying out a morphological dilation of said contour prior to step c2), and that step c2) is carried out on said dilated contour.

5. A computer-implemented method according to claim 1, wherein said at least one segment of said ellipse is fuzzy.

6. A computer-implemented method according to claim 1, wherein step b) further comprises the detection of a local contour orientation for each pixel in said contour.

7. A computer-implemented method according to claim 6, wherein said at least one segment of said ellipse is oriented according to said local contour orientation.

8. A computer-implemented method according to claim 1, wherein, prior to step c4), the following steps are carried out on each one of a plurality of local intensity maxima in said accumulator array:

(i) locating a first position in a circumference centred on each local intensity maximum in the accumulator array, wherein said first position has the highest intensity in said circumference;

(ii) locating a second position in said circumference, wherein said second position has the highest intensity in an arc of said circumference opposite to the first position; and

(iii) weighting the intensity m of the local intensity maximum with a factor W=m²/m_c1·

m_c2, wherein m_c1 is the intensity in said first position, and m_c2, is the intensity in said second position.

9. A computer-implemented method according to claim 1, wherein, prior to step c4), the intensity of each one of a plurality of local intensity maxima in the accumulator array is weighted on the basis of the output of at least one additional head recognition method selected from among a group comprising skin tone recognition, pattern matching, hair recognition, chin detection, hole detection, detection of a connection with a centre of mass or an extremity of said human body, distance to a virtual backbone, upper body axis, and/or previous head position in a video sequence.

10. A computer-implemented method according to claim 1, further comprising, if said source image is a current frame of a video sequence, and a position selected in a preceding frame is occluded in said current frame, disregarding any local intensity maximum and maintaining the head position selected in the precedent frame as corresponding to the head position in the current frame.

11. A non-transitory computer-readable data storage medium containing computer-executable instructions, when executed on a computer, cause the computer to perform a computer-implemented method for recognising a human head in a source image, the source image having depth values for each pixel thereof, the method comprising the steps of:

a) segmenting at least part of a human body from a background in said source depth image;

b) detecting edges of said segmented part of a human body to determine a contour of said part; and

c) determining the position of a human head within the source image;

characterised in that step c) further comprises the steps of;

c1) calculating a depth of said part of a human body in said source image;

c2) calculating a major radius size and a minor radius size of an ellipse corresponding to the size of a human head at said calculated depth in said source image;

c3) generating an accumulator array for at least one segment of an ellipse centred on the positions of a set of contour pixels of said contour, said at least one segment of said ellipse having said major and minor radius sizes; and

c4) selecting a position of a local intensity maximum in said accumulator array that corresponds to the position of a human head in the source image.

12. A computer system programmed for performing a computer-implemented method for recognising a human head in a source image, the source image having depth values for each pixel thereof, the method comprising the steps of:

a) segmenting at least part of a human body from a background in said source depth image;

b) detecting edges of said segmented part of a human body to determine a contour of said part; and

c) determining the position of a human head within the source image;

characterised in that step c) further comprises the steps of;

c1) calculating a depth of said part of a human body in said source image;

c2) calculating a major radius size and a minor radius size of an ellipse corresponding to the size of a human head at said calculated depth in said source image;

c3) generating an accumulator array for at least one segment of an ellipse centred on the positions of a set of contour pixels of said contour, said at least one segment of said ellipse having said major and minor radius sizes; and

c4) selecting a position of a local intensity maximum in said accumulator array that corresponds to the position of a human head in the source image.

13. A computer system according to claim 12 and comprising an imaging device for capturing said source image.

14. A computer system according to claim 13, wherein said imaging device is a 3D imaging device for also capturing said depth.